Page 25 - Read Online
P. 25

Ladeira et al. Microbiome Res Rep 2023;2:9  https://dx.doi.org/10.20517/mrr.2023.01  Page 7 of 15

























                Figure 1. Bifidobacterium-based gut microbiome partitions. (A) Model fit according to BIC; (B) distribution of subjects across the 7
                partitions. Partition #7 consists of subjects for whom no Bifidobacterium was detected (no-Bif); (C) scaled contribution of each
                Bifidobacterium species to each partition. Higher DMM model contributions are associated with a higher relative abundance of a
                particular species. Species are ordered according to hierarchical clustering based on Euclidean distance. Gray indicates alpha parameters
                      -5
                below 10 . More abundant species: B. longum to B. dentium and sub-dominant species: B. moukalabense to B. pseudolongum.


















                Figure 2. Distribution of subjects between Bifidobacterium-based gut microbiome partitions according to metadata (without recent
                antibiotic intake). (A) Age category; (B) westernized lifestyle (adults); (C) health status (adults).

               the cMD (> 1.000 adults) and extracted their Bifidobacterium partitions: (1) 1.098 individuals from the UK
               enrolled in the Personalised Responses to Dietary Composition Trial (PREDICT 1) study ; (2) 1.135
                                                                                               [12]
               participants from the Dutch population-based cohort LifeLines-DEEP ; and (3) 800 individuals from an
                                                                            [18]
               Israeli cohort. All the identified partitions were detected in the three cohorts with different prevalences
               [Supplementary Figure 5]. As for the cMD, the ado_lon and ado_lon_bif partitions were the most prevalent
               (> 50%), and the no_Bif partition was the least prevalent (< 10%). We identified several partitions related to
               Bifidobacterium composition in the adult gut microbiome. These differences were also observed when only
               adults from the cMD were selected [Supplementary Table 1].

               We then used DESeq2 to identify bacterial species for which abundance differed between partitions
               (FDR < 0.05, Wald test) [Figure 3C]. We found that Bifidobacterium partitions enriched in both B. longum
               and B. adolescentis (ado_lon, ado_lon_bif, and lon_ado_cat) shared common, differentially abundant
               bacterial species compared to other partitions (contrasts on the left and right parts of the heatmap). The
               ado_lon, ado_lon_bif, and lon_ado_cat partitions had a lower abundance of E. coli and a higher abundance
               of several butyrate producers (Roseburia faecis, Coprococcus catus, C. eutactus, C. comes, and Eubacterium
   20   21   22   23   24   25   26   27   28   29   30